I think it would be interesting to see if the data on the way fruit flies make decisions could be mapped into the theoretical construct from switching and automata theory called a Stochastic Sequential Machine (SSM). If they can, then it would be instructive to go the next step and explore the possible ways the SSM could be constructed from simpler deterministic and nondeterministic elements. This might give insights into the connectivity, complexity, etc. of the underlying neural networks. This, in turn, might suggest fruitful directions for future experimental research. A paper that gives many ways of constructing SSMs from simpler parts is: "Stochastic Combinational Networks", Information Sciences 16, pp 185-234 (1978).
